I'm a self-taught software engineer from Buenos Aires.

No bootcamp. No university. My grandfather taught me that you can't understand something until you've taken it apart — so I took everything apart.

I started with HTML in 2021. First clients in 2022. First real apps in 2023. By 2024 I was building a full SaaS from scratch, alone.

Then AI arrived and changed the game entirely.

Most devs using AI have one machine, one context, one agent at a time.

I had four machines sitting there — a Mac Mini M4 Pro, two high-end PCs, a laptop. And I kept thinking: what if Tony Stark didn't just have one JARVIS instance? What if all of them worked in parallel?

That's what I built.

ATLAS (Mac Mini M4 Pro)  ──────── orchestrator
    ├── NOVA  (i5 + RTX 3060, Windows)   ──  backend agent
    ├── PIXEL (Ryzen 7 + RTX 3070, Linux) ── frontend agent
    └── NOMAD (Ryzen 7, mobile)           ──  backup / overflow

Four machines. Two Claude accounts. Real-time coordination via SSH + Tailscale. Parallel agents that actually sync between rounds.

No API costs. No cloud dependency. 100% yours.

→ JARVIS — jarvis-490206.web.app/es

The companies don't want self-taught developers. I know because I've sent the applications.

There are thousands of skilled devs locked out of the market — not because they can't build, but because they don't have the right credentials on paper.

JARVIS is my answer to that. If you can't get hired by a team, become the team. The tools exist. You just need to know how to wire them together.

Everything I learned building this is open source.

LocalCenter — A business management SaaS for Argentine companies. AFIP integration, SQL Server, Angular. Full stack, end to end, solo. The kind of project that teaches you what no tutorial ever will.

2021  →  First HTML/CSS/JS. Learning by breaking things.
2022  →  First clients. Real deadlines. Real pressure.
2023  →  TypeScript, databases. First apps people actually used.
2024  →  Full stack SaaS. Building LocalCenter from scratch.
2025  →  Claude Code. AI changes how I work. Forever.
2026  →  Rust. JARVIS. 4 machines. 1 developer. Team output.

I don't start with the stack. I start with the problem.

The stack is a consequence — it adapts to what needs to be built and how fast it needs to exist. I'm not less technical because of this. I'm technical with direction. There's a difference between knowing every API by heart and knowing exactly what to build and why.

With AI, the ceiling for solo developers got a lot higher. But what actually separates people isn't who knows more syntax — it's who can look at a real need, design the right solution, and make it exist. That's closer to an architect than a coder.

No barriers. No "I don't know that language." Just: what needs to exist, and how do I make it exist?

• A solo developer with good tools can out-ship a team with bad ones
• Self-taught isn't a limitation — it's a different kind of education
• The best way to learn something is to build something people actually use
• AI doesn't replace developers. It multiplies the ones who know how to use it.